The large-scale tracer test conducted at the Borden site included detailed three-dimensional plume monitoring, but most previous studies have emphasized the interpretation of the data in a two-dimensional sense. Here, an independent analysis of the fully three-dimensional nature of the plume is developed using a linear spatial interpolation scheme which differs significantly from the gridded higher-order interpolation scheme used in earlier analyses. A detailed examination of the data base reveals significant truncation of the plume during several sampling sessions. This feature explains the low mass estimates obtained during the later sampling sessions. Consequently, there are significant uncertainties associated with the corresponding second-moment estimates. This uncertainty, which calls into question any refined comparisons of second-moment evolution with stochastic theories, has not been recognized in previous interpretations of the second-moment estimates. The estimated vertical macrodispersivity is found to be about twice that attributable to molecular diffusion. The estimated horizontal transverse macrodispersivity is much larger than the vertical macrodispersivity. Temporal variations in the flow appear to provide a plausible explanation of the large horizontal transverse macrodispersivity. Anamolous behavior of the second-moment evolution suggests the role of a large-scale heterogeneity during the later portion of the test, indicating possible mutual interactions between the second-moment components.